FirsthandHealth
medRxiv PreprintsInternational19 November 2026

ClinDataBench: Benchmarking Large Language Model Agents in Clinical Trial Analysis Dataset Construction

This is an official announcement record

Firsthand records what medRxiv Preprints announced and links to the original. The wording below is theirs, not ours.

A new drug reaches patients only after the U.S. Food and Drug Administration (FDA) has reviewed its trial evidence, which every application carries as analysis datasets in the Clinical Data Interchange Standards Consortium (CDISC) Analysis Data Model (ADaM) format that programmers derive by hand from raw trial data, a statistical analysis plan (SAP) and dataset specifications. Coding agents built on large language models (LLMs) could take over this work, and a slow-changing industry will adopt them only with evidence of what they get right and wrong. However, work on large language models (LLM
Source: medRxiv Preprints
Read the official announcement

Opens www.medrxiv.org

More from medRxiv Preprints

This content is for informational purposes only and is not medical advice. It is not intended to diagnose, treat, cure, or prevent any disease. Consult a healthcare professional before starting any supplement, treatment, or program: especially if you are pregnant, nursing, taking medication, or managing a health condition.